Cancer driver gene detection in transcriptional regulatory networks using the structure analysis of weighted regulatory interactions

TL;DR

This paper introduces a novel algorithm that leverages weighted gene regulatory networks and stochastic analysis to identify cancer driver genes, outperforming existing methods in ranking and detection accuracy.

Contribution

The study presents a new network-based algorithm combining topological features and stochastic link-structure analysis for cancer driver gene detection.

Findings

The proposed method achieves higher F-measure scores compared to 23 existing methods.

It effectively ranks genes based on their influence and potential as driver genes.

The algorithm identifies a comprehensive set of candidate driver genes in gene regulatory networks.

Abstract

Identification of genes that initiate cell anomalies and cause cancer in humans is among the important fields in the oncology researches. The mutation and development of anomalies in these genes are then transferred to other genes in the cell and therefore disrupt the normal functionality of the cell. These genes are known as cancer driver genes (CDGs). Various methods have been proposed for predicting CDGs, most of which based on genomic data and based on computational methods. Therefore, some researchers have developed novel bioinformatics approaches. In this study, we propose an algorithm, which is able to calculate the effectiveness and strength of each gene and rank them by using the gene regulatory networks and the stochastic analysis of regulatory linking structures between genes. To do so, firstly we constructed the regulatory network using gene expression data and the list of regulatory interactions. Then, using biological and topological features of the network, we weighted the regulatory interactions. After that, the obtained regulatory interactions weight was used in interaction structure analysis process. Interaction analysis was achieved using two separate Markov chains on the bipartite graph obtained from the main graph of the gene network. To do so, the stochastic approach for link-structure analysis has been implemented. The proposed algorithm categorizes higher-ranked genes as driver genes. The efficiency of the proposed algorithm, regarding the F-measure value and number of identified driver genes, was compared with 23 other computational and network-based methods.